Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

RESEARCH ARTICLE

Preliminary Study of Confounder-Corrected


Fat Fraction and R2* Mapping of Bone
Marrow in Children With Acute Leukemia
Linlin Wang, MD, MSc,1 Dao Wang, MD, PhD,2 Jiao Chen, MD, MSc,2
Mengtian Sun, MD, MSc,1 Dominik Nickel, PhD,3 Stephan Kannengiesser, PhD,3
Feifei Qu, PhD,4 Jingxia Zhu, PhD,4 Cuiping Ren, MD, MSc,1* Yong Zhang, MD, PhD,1* and
Jingliang Cheng, MD, PhD1*

Background: The bone marrow (BM) evaluation of acute leukemia (AL) mainly depends on invasive BM puncture biopsy.
Noninvasive and accurate MR examination technology has potential clinical application value in the BM evaluation of AL
patients. Multi-gradient-echo (MGRE) has been found useful to evaluate changes in BM fat and iron content, but has not
yet been applied in AL.
Purpose: To explore the diagnostic capability of BM infiltration of quantitative BM fat fraction (FF) and R2* values
obtained from a 3D MGRE sequence in children with primary AL.
Study Type: Prospective.
Population/Subjects: Sixty-two pediatric patients with untreated AL and 68 healthy volunteers. AL patients were divided
into acute lymphoblastic leukemia (ALL) (n = 39) and acute myeloid leukemia (AML) (n = 23) groups.
Field Strength/Sequence: 3T, 3D chemical-shift-encoded multi-gradient-echo, T1WI, T2WI, T2_STIR.
Assessment: BM FF and R2* values were assessed by manually drawing regions of interest at the L3, L4, ilium, and 1 cm
below the bilateral trochanter of the femur (upper femur).
Statistical Tests: Independent sample t-tests, variance analysis, Spearman correlation.
Results: BM FF and R2* at L3, L4, ilium, and upper femur, FFtotal and R2*total were significantly lower in the AL than control
group. BM FF did not significantly differ between ALL and AML groups (PL3 = 0.060, PL4 = 0.086, Pilium = 0.179, Pupper
femur = 0.149, and Ptotle = 0.097, respectively). The R2* was significantly lower in ALL group than AML group for L3, L4,
and R2*total. BM FF was moderately positively correlated with R2* in ALL group, and strongly positively correlated in AML
group. Area under the receiver operating characteristic curves showed that BM FF had higher AUC in AL, ALL, and AML
(all AUC = 1.000) than R2* (0.976, 0.996, and 0.941, respectively).
Data Conclusion: MGRE-MRI mapping can be applied to measure BM FF and R2* values, and help evaluate BM infiltration
and iron storage in children with AL.
Evidence Level: 1
Technical Efficacy: 2
J. MAGN. RESON. IMAGING 2023;58:1353–1363.

is increasing yearly.3,4 AL is mainly divided into acute lympho-


A cute leukemia (AL) is among the most common malig-
nant tumors in children and a malignant clonal disease of
the hematopoietic stem cells.1,2 The mortality rate is highest in
blastic leukemia (ALL) and acute myeloid leukemia (AML).
ALL mainly occurs in children and accounts for 80% of all
children with tumor-associated diseases, and the incidence rate leukemias. Currently, leukemia is diagnosed based mainly on

View this article online at wileyonlinelibrary.com. DOI: 10.1002/jmri.28755

Received Jan 5, 2023, Accepted for publication Apr 10, 2023.

*Address reprint requests to: J.C., Zhengzhou, China. E-mail: fccchengjl@zzu.edu.cn, or C.R., Zhengzhou, China. E-mail: fccrencp@zzu.edu.cn, or Y.Z.,
Zhengzhou, China. E-mail: zhy6290@163.com
Linlin Wang and Dao Wang shares co-first authorship.

From the 1MRI Department of the First Affiliated Hospital, Zhengzhou University, Zhengzhou, China; 2Department of Paediatrics of the First Affiliated Hospital,
Zhengzhou University, Zhengzhou, China; 3MR Application Predevelopment, Siemens Healthcare GmbH, Erlangen, Germany; and 4MR Collaboration, Siemens
Healthcare Ltd., Beijing, China

© 2023 International Society for Magnetic Resonance in Medicine. 1353


15222586, 2023, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/jmri.28755 by Manager Information Resources Bond University Library, Wiley Online Library on [14/12/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Journal of Magnetic Resonance Imaging

bone marrow (BM) biopsies and flow cytometry. Although patients. Sixty-two consecutive patients with untreated ALs
BM biopsies are reliable, they are locally invasive. Therefore, confirmed by BM biopsy in our hospital (35 boys and
repeated examinations increase patients’ physical discomfort 27 girls, mean age 8.56  2.46 years, range 5–15 years) were
and psychological fear of the puncture. Thus, noninvasive and included in the study. The control group comprised 68 age-
accurate auxiliary examination techniques are needed to help matched healthy child volunteers (47 boys and 21 girls, mean
assess the BM condition in children with leukemia to reduce age 8.19  2.60 years, range 5–15 years) (Fig. 3a,b). Refer-
or prevent repeated BM punctures. ring to the 2016 World Health Organization hematopoietic
Fat assessment can show whether the tumoral lesion and lymphoid tumor classification criteria, children with AL
invades the BM and help predict the benign or malignant were divided into the ALL (n = 39, 22 boys and 17 girls,
and extent of BM lesions; moreover, it can prevent >60% of mean age 7.72  2.58 years) (Fig. 3e,f) and AML (n = 23,
biopsies in patients without tumors.5,6 MRI is a noninvasive 13 boys and 10 girls, mean age 9.00  2.47 years) (Fig. 3c,d)
imaging technique that enables assessment of fatty tissue, groups.20 The AML group included types M0 (n = 1), M2
especially in the liver rapid developments in dual-echo (n = 11), M3 (n = 8), M4 (n = 2), and M5 (n = 1).
Dixon-type water-fat separation and multi-echo methods have
enabled MR-based assessment of fat to become widely used
in clinical practice.7–9 Early applications to BM quantification
used the three-echo approach to correct the T2* effect.10
State-of-the-art implementations are typically based on 3D
chemical-shift-encoded multi-gradient-echo (MGRE) acquisi-
tions with simultaneous data fitting for “proton density fat
fraction” (PDFF) and transverse relaxation rate (R2*) while
accounting for confounding effects like T1 relaxation and the
spectral complexity of fat.11 Compared with the reference
standard multi-echo single-voxel MR spectroscopy (MRS),
studies have confirmed that the fat content obtained by
MGRE MRI shows better correlation and consistency.12–14
Previous research has also reported the accuracy of R2* mea-
surements (or its inverse, T2*) for the iron content in
abdominal organs and models experiment.15–17 MGRE MRI
allows obtaining quantitative BM fat fraction (FF) and R2*
values with noninvasive imaging.
AL mainly invades the BM and affects the hematopoi-
etic system, especially the red BM, which can cause the fat
content in the BM to decrease.18 In the pathological state,
the yellow BM may revert to red BM, whose content will
then also decrease. Additionally, the iron content in the BM
will change with the course and blood transfusion treatment
that would affect the prognosis.19 The use of R2* values to
determine the BM iron content in patients with primary AL
is not commonly reported in the literature, and BM FF is
rarely reported in pediatric patients with primary AL.
Therefore, the goal of this study was to quantitatively
measure BM FF and R2* values in the lower lumbar verte-
brae, pelvis and upper femur of children with MGRE; to
investigate myelogram, hemogram, FF and R2* correlations FIGURE 3: (a and b) Male, 12 years, FF and R2* picture of a
and to evaluate the diagnostic ability for BM infiltration of control, comparison with muscle tissue, both FF and R2*
pictures show high signals in bone marrow; the former has
this technique for AL.
higher signal. (c and d) Female, 12 years, FF and R2* picture of
AML patient, FF shows that the bone marrow signal is diffusely
reduced and equal to the muscle tissue signal, R2* picture
Patients and Methods shows bone marrow signal is close to muscle tissue. (e and f)
Female, 7 years, ALL patient, on FF and R2* map, bone marrow
Participants
signal performance is consistent with AML patients (the oval
The institutional review board of our hospital approved the represents the area of interest, at the L3, L4, ilium, and upper
study and written informed consent was obtained from all femur levels).

1354 Volume 58, No. 5


15222586, 2023, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/jmri.28755 by Manager Information Resources Bond University Library, Wiley Online Library on [14/12/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Wang et al.: Preliminary Study of Confounder-Corrected Fat Fraction

Myelograms were observed for the percentages of primitive manually delineated regions of interest (ROI). The ROIs
and immature cells. Inclusion criteria for the control subjects were manually drawn on the FF maps (Fig. 3a,c,e). At the
were no abnormalities in the peripheral blood; no blood or same time, the ROIs were copied to the R2* relaxation maps
BM system disease. Four patients were excluded because of (Fig. 3b,d,f). The ROIs in the medullary cavity areas of the
inadequate examination results or motion artifacts. median levels of the L3 and L4 vertebrae, the bilateral ilium,
and the bilateral 1 cm under the bilateral lesser trochanter of
MRI Protocol the upper femur were drawn in turn and using conventional
An MRI scan was performed on all participants using a 3T coronal T1WI, T2WI, and T2_STIR sequences as references.
system (MAGNETOM Prisma, Siemens Healthcare, Quantitative BM FF and R2* values were obtained as mean
Erlangen, Germany) with one (for patients 5–12 years) or values over the ROIs from the MGRE FF and R2* map,
two (for patients 13–15 years) 18-channel phased-array body respectively. The ROIs were avoided the bone cortex and ter-
coils in combination with a 32-channel phased-array spine minal plate of the vertebral body as much as possible. Care
coil. MRI examinations were performed within 48 hours of was taken to ensure that the size and position of the ROIs
the BM biopsies. All participants were examined in supine were as same or similar as possible because the ROI may vary
position, with both feet together. The coil covered the lower slightly owing to the influence of the position of the medul-
lumbar vertebrae, pelvis, and upper femur. lary cavity. Each ROI was measured three times, and the
The MRI protocol consisted on: the multi-gradient- average was taken as the final result for any site. In this arti-
echo prototype sequence (MGRE), T1-weighted imaging cle, FF and R2* values of ilium and 1 cm under the bilateral
(T1WI), T2-weighted imaging (T2WI) and T2-weighted lesser trochanter of the upper femur were the average value
short-TI inversion recovery (T2_STIR) sequences. Table 1 measured from the bilateral sites. Finally, results of the indi-
lists all MRI sequence parameters. MGRE used coronal slice vidual regions of interest were averaged together and reported
orientation, a lower parallel imaging acceleration factor, mul- as FFtotal and R2*total.
tiple signal averages, no breath-holding, and low-rank den-
oising of the source images.21 This prototype sequence can Statistical Analysis
collect high spatial resolution images that generates FF with SPSS 21.0 software (IBM Corp, Armonk, NY, USA) was
comparable accuracy of spectroscopy. Quantitative FF and used for statistical analyses. Continuous values were expressed
R2* maps were automatically generated inline after data as mean  standard deviation. Independent sample t-tests
acquisition from the MGRE data by a confounder-corrected were used to compare the BM FF and R2* values,
multi-step fitting approach.22 myelogram, lymphocyte percentage, and neutrophil percent-
age between groups. The FF and R2* values between posi-
Quantitative Image Assessment tions were compared by variance analysis. Pearson correlation
After scanning, the MGRE was automatically reconstructed analysis was performed for the BM FF, R2* value, and labo-
to generate the FF map and R2* relaxation maps, and all ratory indicators that were normally distributed. Spearman
images were transferred to a processing workstation (syngo. correlation analysis was used for non-normal distributions.
via, Siemens Healthcare, Erlangen, Germany) for analysis. A Receiver operating characteristic (ROC) curves were used to
radiologist (L.W.) involved in the MRI diagnosis for 12 years analyze the diagnostic capability of the BM FF and R2*

TABLE 1. MRI scan parameters

GRAPPA
Resolution (mm3) TR (msec) TE (msec) Averages Factor Flip Angle ( ) Orientation Scan Time
MGRE 1.1  1.1  3.5 9.15 1.07, 2.24, 8 2 4 Coronal 2 minutes 18 s
3.41, 4.58,
5.75, 6.92
T1WI 0.8  0.8  5.0 550 21 1 3 140 Coronal 1 minute 12 s
T2WI 0.8  0.8  4.0 4000 86 1 2 150 Coronal 1 minute 22 s
T2_STIR 1.4  1.4  4.0 6000 46 1 2 150 Coronal 1 minute 20 s

MGRE = multi-gradient-echo; TR = repetition time; TE = echo time; TIWI = T1-weighted imaging; T2WI = T2-weighted imaging;
T2_STIR = T2-weighted short-TI inversion recovery; GRAPPA = Generalized Autocalibrating Partially Parallel Acquisitions.

November 2023 1355


15222586, 2023, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/jmri.28755 by Manager Information Resources Bond University Library, Wiley Online Library on [14/12/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Journal of Magnetic Resonance Imaging

values, which was performed with the mean BM FF and R2* Comparison of BM FF and R2* Values Between the
values obtained from bilateral iliac bones of AL, ALL, and ALL and AML Groups
AML patients at common puncture sites. The ROC AUC Age did not significantly differ between the ALL (n = 39)
was assessed with a confidence interval (IC) of 95%. P < 0.05 and AML (n = 23) groups (P = 0.060). Both groups showed
was considered statistically significant. inconsistent BM FF and R2* values in whether there is a sta-
tistical difference. These values for the L3, L4, ilium, upper
femur, as well as FFtotal and R2*total were lower in the ALL
Results group than in the AML group, but BM FF did not signifi-
Comparison of BM FF and R2* Values Between the cantly differ (PL3 = 0.060, PL4 = 0.086, Pilium = 0.179,
AL and Control Groups Pupper femur = 0.149, and Ptotle = 0.097, respectively). The
Age did not significantly differ between the AL and control R2* value was significantly lower in the ALL group than in
groups (P = 0.412). The BM FF and R2* values at the L3, the AML group for L3, L4, and R2*total. R2* values in the
L4, ilium, upper femur, as well as FFtotal and R2*total levels, ilium and upper femur were slightly lower in the ALL group
were significantly lower in the AL group than in the controls than in the AML group, but not significantly (P = 0.052 and
(Table 2). P = 0.646, respectively) (Table 3).
The BM FF in the control group gradually increased BM FF did not differ significantly for the L3, L4, ilium
from the L3, L4, and ilium to the upper femur levels, and and femur in the ALL group (P = 0.066). In the AML
R2* values also showed a gradually increased from the L3, L4 group, the L3 and L4 BM FFs were lower than those of
to ilium except for upper femur levels. However, the BM FF the ilium and femur, but not significantly (P = 0.568).
and R2* values between L3 and L4 did not significantly dif- However, the R2* values of the upper femur in both
fer (P = 0.246 and P = 0.589, respectively). Significant dif- groups were significantly lower than those of the L3, L4,
ferences were found between the L3 and ilium, L3 and upper and ilium. No significant difference was found between
femur, L4 and ilium, L4 and upper femur, and ilium and L3 and L4 (PALL = 0.900; PAML = 0.926), L3 and the
upper femur. R2* values were significantly lower in the upper ilium (PALL = 0.329; PAML = 0.902), and L4 and the
femur than in the L3, L4, and ilium. ilium (PALL = 0.394; PAML = 0.976) (Table 3).
In the AL group, the FF was lower for the L3 and L4
vertebrae than for the ilium and upper femur, but not signifi- Correlation Analysis of BM FF and R2* Value in the
cantly (P = 0.083). The R2* values were significantly lower Control, AL, ALL, and AML Groups
in the upper femur than at the L3, L4, and ilium levels Correlation analysis was performed on the mean values of the
(Table 2). FF and R2* values in the control group and patients, and the

TABLE 2. Clinical data and imaging characteristics of the control and AL groups

(FF [%]) (R2* [s 1])


Variable Control AL Group P Value Control AL Group P Value
Participants n = 68 n = 62 n = 68 n = 62
Sex (M/F) 47/21 35/27 47/21 35/27
Age (years) 8.56  2.46 8.19  2.60 0.412 8.56  2.46 8.19  2.60 0.412
Target location
L3 40.19  9.43 2.71  2.93 <0.05 147.62  25.21 87.37  30.27 <0.05
L4 41.85  9.76 2.99  3.28 <0.05 149.73  23.84 87.45  26.52 <0.05
Ilium 54.19  5.50a 3.92  4.65 <0.05 164.23  18.36b 90.19  26.77 <0.05
Femur 62.37  8.01 a
4.38  5.18 <0.05 96.73  23.23 b
57.32  28.92 c
<0.05
Total 49.65  6.68 3.50  3.55 <0.05 139.58  15.84 80.58  23.53 <0.05

M = Male; F = Female; FF = fat fraction; R2* = R2* values; AL = acute leukemia.


a
Significant differences in bone marrow FF between L3 and ilium, L3 and femur, L4 and ilium, L4 and femur, ilium and femur.
b
Significant differences in R2* values between L3 and ilium, L3 and femur, L4 and ilium, L4 and femur, ilium and femur.
c
Significant differences in R2* values between L3 and femur, L4 and femur, and ilium and femur.

1356 Volume 58, No. 5


15222586, 2023, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/jmri.28755 by Manager Information Resources Bond University Library, Wiley Online Library on [14/12/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Wang et al.: Preliminary Study of Confounder-Corrected Fat Fraction

TABLE 3. Clinical data and imaging characteristics of ALL and AML patients

Result (FF [%]) Result (R2*[s 1])


Variable ALL AML P Value ALL AML P Value
Participants 39 23 39 23
Sex (M/F) 22/17 13/10 22/17 13/10
Age (years) 7.72  2.58 9.00  2.47 0.060 7.72  2.58 9.00  2.47 0.060
Target location
L3 2.10  1.39 3.75  4.33 0.060 79.30  22.84 101.06  36.43 0.015
L4 2.37  2.22 4.03  4.41 0.086 79.96  18.01 100.16  33.48 0.012
Ilium 3.18  2.75 5.19  6.65 0.179 84.49  20.85 99.87  32.87 0.052
Femur 3.54  3.85 5.81  6.73 0.149 56.01  30.16a 59.54  27.19a 0.646
Total 2.79  1.99 4.69  5.07 0.097 74.94  17.79 90.16  28.92 0.029

M = Male; F = Female; ALL = acute lymphoblastic leukemia; AML = acute myeloid leukemia; FF = fat fraction; R2* = R2* values.
a
Significant differences in R2* values between L3 and femur, L4 and femur, and ilium and femur.

mean values of BM FF and R2* values are derived from the TABLE 4. Spearman correlations between AL patient
sum of L3, L4, ilium, and upper femur, respectively. No cor- bone marrow FF and R2* values
relation was found between BM FF and R2* values in the
control group (R = 0.134, P = 0.277). BM FF was moder- Parameters n R P Value
ately positively correlated with R2* values in all patients FFcon and R2*con 68 0.134 0.277
(R = 0.499), and in the ALL group (R = 0.476). FF and FFAL and R2*AL 62 0.499 0.000*
R2* values were strongly positively correlated in the AML
group (R = 0.615; Table 4; Fig. 1a–d). FFALL and R2*ALL 39 0.472 0.002*
FFAML and R2*AML 23 0.615 0.002*

AL = acute leukemia; ALL = acute lymphoblastic leukemia;


Comparison of Laboratory Indicators and AML = acute myeloid leukemia; FFcon = fat fraction of control
Correlation Analysis Between the ALL and AML groups; R2*con = R2* value of control groups; FFAL = fat frac-
Groups tion of AL; R2*AL = R2* value of AL; FFALL = fat fraction of
The proportions of primitive and naive leukemia cells on the ALL; FFAML = fat fraction of AML; R2*ALL = R2* value of
ALL; R2*AML = R2* value of AML.
myelogram and the percentage of peripheral blood lympho-
cytes were significantly higher in the ALL group than in the
AML group. The percentage of peripheral blood neutrophils
was significantly lower in the ALL group than in the AML respectively; P = 0.106 and P = 0.067, respectively;
group (Table 5). Table 6).
No significant correlation was found between the BM
FF and R2* values and the myelogram and percentage of Diagnostic Performance Comparison of BM FF and
peripheral blood lymphocytes and neutrophils in the ALL R2*Values in AL, ALL, and AML Patients
group (P = 0.271, 0.925, 0.864, 0.469, 0.357, and 0.175, BM FF and R2* values had high diagnostic performance for all
respectively). In the AML group, BM FF was moderately neg- patients, with BM FF performing the best (Table 7; Fig. 2a–c).
atively correlated with myelogram and neutrophil percentage The area under the curve (AUC) of the AL (1.000;
(R = 0.595 and R = 0.420, respectively) and significantly 95% confidence interval [95% CI]: 0.972–1.000) was slightly
positively correlated with lymphocyte percentage higher than that of the ALL (1.000; 95% CI: 0.966–1.000)
(R = 0.676). The AML R2* value was moderately positively and AML (1.000; 95% CI: 0.960–1.000) BM FFs. The 95%
correlated with the lymphocyte percentage (R = 0.517), and CIs of the R2* values were 0.932–0.995, 0.959–1.000, and
no significant correlation was found between the myelogram 0.871–0.980 for AL, ALL, and AML, respectively. The sensi-
and neutrophil percentage (R = 0.346 and R = 0.388, tivity of the BM FF for all patients was 100%, which was

November 2023 1357


15222586, 2023, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/jmri.28755 by Manager Information Resources Bond University Library, Wiley Online Library on [14/12/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Journal of Magnetic Resonance Imaging

FIGURE 1: Correlation analysis of the BM FF and R2* values in the control (a), AL (b), ALL (c), and AML (d) groups and of the mean of
the sum of the FF and R2 values. AL = acute leukemia; ALL = acute lymphoblastic leukemia; AML = acute myeloid leukemia.

TABLE 5. Comparison of laboratory indicators between ALL and AML groups

Myelogram (%) Lymphocyte percentage (%) Neutrophil percentage (%)


Parameters n Mean  SD p Value Mean  SD P Value Mean  SD P Value
ALL 39 88.14  16.29 0.002 76.05  12.85 <0.05 16.19  11.41 <0.05
AML 23 71.19  22.47 40.74  24.33 40.39  25.49

ALL = acute lymphoblastic leukemia; AML = acute myeloid leukemia.

higher than that of the corresponding BM R2* value (88.71, Discussion


94.87, and 78.26, respectively). The specificity of both In this study, we investigated the potential diagnostic ability
parameters was the same (100%). The maximum reference of MGRE in diffuse BM infiltration of AL in children. We
values of the BM FF in the AL, ALL, and AML patients were found that quantitative BM FF and R2* values can accurately
22.38%, 12.81%, and 22.38%, respectively, and the maxi- evaluate the changes in fat content and the iron content in
mum reference values of the R2* values were 126.41, 111.7, the BM of AL. Compared with normal children, quantitative
and 126.41 s 1, respectively (Table 7; Fig. 2a–c). BM FF significantly decreased in the BM of children with

1358 Volume 58, No. 5


15222586, 2023, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/jmri.28755 by Manager Information Resources Bond University Library, Wiley Online Library on [14/12/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Wang et al.: Preliminary Study of Confounder-Corrected Fat Fraction

TABLE 6. Spearman correlations between laboratory indicators and FF and R2* values of AL patients

Correlation Correlation
Variable R P Value Variable R P Value
FFALL and M 0.181 0.271 FFAML and M 0.595 0.003*
FFALL and LP 0.010 0.952 FFAML and LP 0.676 0.000*
FFALL and NP 0.028 0.864 FFAML and NP 0.420 0.046*
R2*ALL and M 0.120 0.469 R2*AML and M 0.346 0.106
R2 ALL and LP 0.152 0.357 R2*AML and LP 0.517 0.011*
R2*ALL and NP 0.222 0.175 R2*AML and NP 0.388 0.067

AL = acute leukemia; ALL = acute lymphoblastic leukemia; AML = acute myeloid leukemia; FFALL = fat fraction of ALL;
FFAML = fat fraction of AML; R2*ALL = R2* value of ALL; R2*AML = R2*value of AML; M = myelogram; LP = lymphocyte per-
centage; NP = neutrophil percentage.

AL, especially in ALL patients. The BM R2* value was signif- lesser trochanter were significantly lower than those in healthy
icantly correlated with BM FF. In addition, we also found children at the same age, but no significant difference was
that ALL and AML showed significant differences in found between ALL and AML groups at L3, L4, bilateral
Myelogram, lymphocyte percentage (LP), and neutrophil per- ilium and 1 cm under the bilateral lesser trochanter. Normal
centage (NP). In contrast, the BM FF of AML patients tissue is replaced by tumor cells because of the abnormal pro-
showed a moderate or obvious correlation with Myelogram, liferation of primitive and immature cells in BM hematopoi-
LP, and NP. The R2* value showed a moderate correlation etic tissue.25 When leukemia cells infiltrate BM, the yellow
with LP, which differed from ALL. Combined with the AUC BM reverts to red BM in the pathological state, and when
of ROC, BM FF, and R2* values have high diagnostic per- normal BM is replaced or reversed, the proportion of water
formance for patients with AL, ALL, and AML. These results and fat in BM changes.26 Different degrees of leukemia cells
may be helpful for the quantitative evaluation of diffuse BM were distributed in the vertebral body, pelvis, and upper
infiltration in AL patients with quantitative FF and R2* femur with rich red BM, and a decreased fat content in the
values. BM resulted in a significant decrease in the BM FF. The
Human BM is divided into hematopoietic red BM and decrease of fat content in BM is shown by obvious signal
fatty yellow BM. The compositions of red and yellow BM reduction in FF map (Fig. 3c,e), and the BM signal of normal
differ as follows: red marrow contains 40% water, 40% fat, children is relatively high (Fig. 3a). Nguyen et al reported
and 20% protein; yellow marrow contains 15% water, 80% that the decrease in leukemia cells is related to an increase in
fat, and 5% protein.23 The distribution of red and yellow the area of BM fat.27 Compared with the control group, the
BM changes dynamically; adipose tissue gradually replaces the areas of BM occupied by adipocytes were significantly
hematopoietic tissue in long bones from shortly after birth to reduced in patients with ALL. Our results are consistent with
childhood, becoming yellow BM and reaching the state of those of Nguyen et al in that the BM FFs of the L3, L4,
adult BM at 25 years of age.24 Consequently, children’s BM ilium, and upper femur in AL patients were generally
undergoes rapid age-related changes, and the fat content in reduced. The R2* values were significantly lower in the AL
the BM and the FF are lower than those in adults. To avoid group than in the controls. In AL, the decreased iron content
the influence of age in this study, the control group com- in the BM may be caused by the special iron intake mecha-
prised age-matched, healthy children. Consequently, it is nism to obtain iron from the host to meet the abnormal pro-
important that in this study, age did not significantly differ liferation needs of leukemia cells. Regarding the significance
between the patients and controls. of BM iron content in patients with AL, research suggests
BM infiltration in leukemia first occurs in the red BM, that reduced BM iron storage can increase the response of
which is related to the origin of leukemic cells in the red BM patients with ALL.19 Iron overload (grade ≥3) will increase
and the abundant blood supply.18 Therefore, in this study, patients’ tolerance to treatment and BM recurrence risk.19
the vertebral body, ilium, and upper femur with rich red BM Further studies of the BM iron reduction mechanism in
were selected as the target sites. The FFs in children with AL patients with AL should provide new treatment options and
at the L3, L4, bilateral ilium and 1 cm under the bilateral methods of prognosis evaluation for patients with leukemia.

November 2023 1359


15222586, 2023, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/jmri.28755 by Manager Information Resources Bond University Library, Wiley Online Library on [14/12/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Journal of Magnetic Resonance Imaging

The BM distributions also differed in the control group.

AL = patients with acute leukemia; ALL = patients with acute lymphoblastic leukemia; AML = patients with acute myeloid leukemia; FF = bone marrow fat fraction of iliac bone;
0.941 (0.871–0.980)
Healthy children showed higher FFs in the upper femur,

78.26 (56.3–92.5)
100 (94.7–100)
followed by the ilium, and finally the L3 and L4. FFs did not

1
≤126.41 s
<0.001
significantly differ between the L3 and L4 vertebrae; however,

0.783
R2*
significant differences were found between the L3 and ilium,

R2* = bone marrow R2* value of iliac bone; sensitivity and specificity rows are 95% confidence intervals. AUC = area under the ROC curve; MCV = maximum reference value.
L3 and upper femur, L4 and ilium, L4 and upper femur, and
ilium and upper femur. This distribution is consistent with
AML

the normal growth patterns of children because BM transfor-


mation occurs earlier and faster in children’s long bones than
1.000 (0.960–1.000) in the central axis and pelvis, although the pelvic bone also
100 (85.2–100)
100 (94.7–100)

undergoes transformation, but it is slower than that of the

≤22.38%
<0.001
1.00
femur and occurs earlier than that of the vertebrae.23 There-
FF

fore, the fat content is more abundant in the long bones than
in the central axis and pelvis, and the FF is higher in the
upper femur, followed by the pelvis, L3, and L4 vertebrae.
The R2* value of the control group increased gradually at the
0.996 (0.959–1.000)

L3, L4, and ilium, which was consistent with the perfor-
94.87 (82.7–99.4)
100 (94.7–100)

mance of the BM FF, but the level of the upper femur


1

<0.001
≤111.7 s

decreased, which was inconsistent with the BM FF perfor-


0.949
R2*

mance. The reasons for this require further investigation. The


increasing and decreasing trend of the R2* values from the
TABLE 7. Diagnostic performance of iliac bone marrow FF and R2*value for AL, ALL, AML patients

L3, L4, ilium and upper femur were consistent with the FF
ALL

in the AL group. The change in the R2* value was accompa-


nied by a change in the BM FF, which is consistent with the
1.000 (0.966–1.000)

results of the liver study by Bashir et al.28 It has to be noted


100 (91.0–100)
100 (94.7–100)

that the signal model for the (simultaneous) parameter fitting


≤12.81%
<0.001
1.00
FF

of FF and R2* used in this study was primarily designed for


quantifying liver fat, so it is possible that some correlation
between the two parameters is caused by inexact modeling of
the potentially different signal evolution of BM compared
with liver parenchyma. However, the magnitude of change in
0.976 (0.932–0.995)

FF and R2* makes it likely that the observed correlation has


88.71 (78.1–95.3)
100 (94.7–100)

a primarily physiological reason.


≤126.41 s
<0.001
0.887

Comparisons of the BM FF and R2* values between ALL


R2*

and AML groups showed that the R2* values changed as the
BM FF changed. The FF and R2* values in the ALL group
were generally lower than those in the AML group at the L3,
L4, ilium, upper femur, and overall levels, which may be related
AL

to the different proportions of primary and immature cells in


1.000 (0.972–1.000)
100 (94.2–100)
100 (94.7–100)

the BM of patients with ALL and AML. The proportion of


≤22.38%

abnormal proliferation of primary and immature leukemia cells


<0.001
1.00
FF

was higher in the ALL group than in the AML group; thus, the
BM FF in the ALL group was generally lower than that in the
AML group, which is consistent with the pathological manifesta-
tions of leukemia in the BM. However, no significant differences
were found in the FFs between the groups. The R2* value of
Sensitivity (95% CI)
Specificity (95% CI)

the ALL group was significantly lower than that of the AML
group at the L3, L4, and total levels. This was consistent with
AUC (95% CI)
Youden index J

the BM FF but appears to be more sensitive than the FF


response. However, no significant differences were found
Variable

P value

between the two groups at the ilium and upper femur levels,
MCV

but the R2* values were lower at the upper femur levels than at
the L3, L4 levels and ilium for the ALL and AML groups.

1360 Volume 58, No. 5


15222586, 2023, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/jmri.28755 by Manager Information Resources Bond University Library, Wiley Online Library on [14/12/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Wang et al.: Preliminary Study of Confounder-Corrected Fat Fraction

FIGURE 2: ROC curve analysis of BM FF and R2* values in AL, ALL, and AML patients. (a) ROC curves of FF and R2* value of AL
patients. (b) ROC curves of FF and R2*value of ALL patients. (c) ROC curves of FF and R2* value of AML patients. Diagonal line
denotes reference. AL = acute leukemia; ALL = acute lymphoblastic leukemia; AML = acute myeloid leukemia; FF = fat fraction;
BM = bone marrow; ROC = Receiver operating characteristic.

Comparing the different levels in the ALL and AML positively correlated in patients with AL, ALL. The R2* value
groups showed no significant differences in BM FFs between decreased as the BM FF decreased, which was consistent with
groups at any level. The differences in BM FF between levels the results of Bashir et al.28 Compared with the AL and ALL
was less obvious in the ALL group than in the AML group. groups, the correlation was higher between the BM FF and
Specifically, in the AML group, the FFs of the L3 and L4 ver- R2* values in the AML group.
tebrae were lower than those of the ilium and upper femur, Myelogram analysis showed that the proportion of pri-
but these levels did not significantly differ in the ALL group. mary and immature leukemia cells was significantly higher in
Therefore, compared with AML, we believe that ALL has the ALL group than in the AML group, which explains why
higher invasion and spread characteristics, and the area of the BM FF at each site was lower in the ALL group than in
BM fat decreases more significantly and is more obvious in the AML group. The BM FF in the AML group was moder-
the ilium and upper femur. This may be related to the pro- ately negatively correlated with the myelogram, but the R2*
portion of primitive and immature cells, biological character- value was not, indicating that the BM FF in the AML group
istics, and different chemical compositions of the cells.18 The decreased as the numbers of primitive and immature cells
changes in R2* values between levels were also generally con- increased. No significant correlation was found between the
sistent with the BM FFs. The R2* values were generally lower myelogram and FF or myelogram and R2* value in the ALL
in the ALL group than in the AML group, and in the ALL group.
and AML groups, the R2* values were significantly lower in There is a significant difference between ALL and AML
the upper femur than in the L3, L4, and ilium. The reasons groups in the lymphocyte percentage and neutrophil percent-
for the lower R2* values in the upper femur are unclear and age in peripheral blood, which may be different from differ-
require further study. Due to anemia, leukemia patients often ent subtypes of leukemia in the type, differentiation, chemical
need blood transfusion treatment. Frequent or massive blood composition, and proliferation rate of tumor cells. ALL origi-
transfusion treatment can cause systemic iron overload. Some nates from B-line or T-line cells of lymphocytes and prolifer-
published works reported that excessive iron can damage hema- ates abnormally in BM. AML is a malignant myeloid
topoiesis by inducing apoptosis and cell cycle arrest, leading to hematopoietic stem/progenitor cells disease, characterized by
a decline in the ratio and clonogenic function of hematopoietic abnormal proliferation of primitive and juvenile myeloid cells
stem and progenitor stem cells.29,30 Brissot et al mentioned in in BM and peripheral blood. In the AML group, the labora-
the study that excessive iron in leukemia patients has harmful tory study showed that the BM FF and R2* values were sig-
effects on both hematopoietic function and prognosis, espe- nificantly correlated with the percentage of lymphocytes, and
cially after hematopoietic stem cell transplantation.31 Moafi the BM FF was significantly negatively correlated with the
et al reported that the increase of BM iron storage (BMIS) at percentage of neutrophils, which can provide additional infor-
the end of the first year of treatment was related to the inci- mation related to the BM fat content and iron storage.
dence of drug resistance and the risk of relapse.19 Therefore, The results of this study suggest that the obtained FF
MGRE could also be a useful tool to monitor the variation of and R2* values, especially the FF, are the best parameters for
iron content in BM during the treatment of leukemia. distinguishing between normal hematopoietic BM and BM
Correlation analysis showed that the BM FF and R2* infiltration in leukemia, consistent with previously reported
values were correlated to varying degrees and were moderately results.32 The literature reports that the FF has good

November 2023 1361


15222586, 2023, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/jmri.28755 by Manager Information Resources Bond University Library, Wiley Online Library on [14/12/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Journal of Magnetic Resonance Imaging

diagnostic performance for differentiating benign and malig-


nant BM lesions such as spinal Schumacher’s nodules from Acknowledgments
metastatic tumors.33 Schmeel et al also confirmed that PDFF This study was supported by a public service platform for
has high diagnostic accuracy, sensitivity, and specificity in dif- artificial intelligence screening and auxiliary diagnosis for the
ferentiating benign and malignant BM lesions through ROC medical and health industry, Ministry of Industry and Infor-
curve analysis.34 Disler et al reported that chemical mation Technology of the People’s Republic of China
shiftencoded MRI can help predict the likelihood of neoplas- (2020-0103-3-1).
tic or nonneoplastic lesions with high diagnostic perfor-
mance.35 Similarly, R2* value can accurately evaluate the References
iron content in BM. Clinically, the evaluation of the iron 1. Wojcicki AV, Kasowski MM, Sakamoto KM, Lacayo N. Metabolomics in
content in patients with AL is usually performed on the labo- acute myeloid leukemia. Mol Genet Metab 2020;130(4):230-238.
ratory index serum ferritin, which is considered the most 2. Juliusson G, Hough R. Leukemia. Prog Tumor Res 2016;43:87-100.
commonly used parameter.36 However, serum ferritin is also 3. Ceppi F, Cazzaniga G, Colombini A, Biondi A, Conter V. Risk factors for
an indicator of inflammation and acute phase reaction, and relapse in childhood acutely mphoblastic leukemia: Prediction and pre-
has a poor correlation with the actual iron content in the vention. Expert Rev Hematol 2015;28(1):57-70.

body.36,37 Therefore, inflammatory infection, tumors and 4. Chang JS, Selvin S, Metayer C, Crouse V, Golembesky A, Buffler PA.
Parental smoking and the risk of childhood leukemia. Am J Epidemiol
autoimmune disease can affect the measured value of serum 2006;163(12):1091-1100.
ferritin. Relevant researchers reported that MRI can achieve
5. Zampa V, Cosottini M, Michelassi C, et al. Value of opposed-phase
high sensitivity and specificity in the analysis of liver iron gradient-echo technique indistinguishing between benign and malig-
content, and it is considered the best surrogate marker of nant vertebral lesions. Eur Radiol 2002;12(7):1811-1818.

total body iron content.38,39 6. Kohl CA, Chivers FS, Lorans R, Roberts CC, Kransdorf MJ. Accuracy of
chemical shift MR imaging in diagnosing indeterminate bone marrow
lesions in the pelvis: Review of a single institution’s experience. Skeletal
Radiol 2014;43(8):1079-1084.

Limitations 7. Berglund J, Ahlström H, Johansson L, Kullberg J. Two-point Dixon


method with flexible echo times. Magn Reson Med 2011;65(4):994-
The first one is the sample size of the study is small. The 1004.
smaller sample size may cause some issues. First, the lower 8. Eggers H, Brendel B, Duijndam A, Herigault G. Dual-echo Dixon imag-
parameter values obtained in some patients may have ing with flexible choice of echo times. Magn Reson Med 2011;65(1):
96-107.
affected the results of the correlation analysis. Second, many
factors can cause changes in BM FF and R2* values in ALL 9. Reeder SB, Cruite I, Hamilton G, Sirlin CB. Quantitative assessment of
liver fat with magnetic resonance imaging and spectroscopy. J Magn
patients, and the increased proportion of primitive and Reson Imaging 2011;34(4):729-749.
immature cells in the BM is only one factor. At the last, the 10. Kim YP, Kannengiesser S, Paek MY, et al. Differentiation between focal
results suggest that local pathological samples from BM malignant marrow-replacing lesions and benign red marrow deposition
of the spine with T2*-corrected fat-signal fraction map using a three-
biopsies may not represent the overall situation of the BM echo volume interpolated breath-hold gradient echo Dixon sequence.
changes in these patients. Therefore, whether the Korean J Radiol 2014;15(6):781-791.
myelogram and BM FF and R2* values are correlated in 11. Karampinos DC, Melkus G, Baum T, Bauer JS, Rummeny EJ, Krug R.
ALL requires further research with a larger sample. The sec- Bone marrow fat quantification in the presence of trabecular bone: Ini-
tial comparison between water-fat imaging and single-voxel MRS.
ond one is that the age range of the patient cohort is quite Magn Reson Med 2014;71(3):1158-1165.
large. Future studies should comprise subgroups with differ-
12. Lee SH, Yoo HJ, Man YS, et al. Fat quantification in the vertebral body:
ent age ranges so that FF and R2* may be more sensitive Comparison of modified Dixon technique with single-voxel magnetic
and accurate in the diagnosis of AL, ALL and AML in those resonance spectroscopy. Korean J Radiol 2019;20(1):126-133.

subgroups. The third limitation is that the lack of compari- 13. Vucht NV, Santiago R, Lottmann B, et al. The Dixon technique for MRI
of the bone marrow. Skeletal Radiol 2019;48(12):1861-1874.
son of BM changes before and after treatment, and a lack of
prognosis evaluation. Future studies should include follow- 14. Li G, Zheng X, Chen Y, et al. Longitudinal assessment of marrow fat
content using three-point Dixon technique in osteoporotic rabbits.
up of the patients’ BM changes to assist in clinical pre- Menopause 2016;23(12):1339-1344.
treatment evaluations, monitoring of treatment effects, and 15. Serai SD, Fleck RJ, Quinn CT, Zhang B, Podberesky DJ. Retrospective
prognostic evaluations. comparison of gradient recalled echo R2* and spin-echo R2 magnetic
resonance analysis methods for estimating liver iron content in children
and adolescents. Pediatr Radiol 2015;45(11):1629-1634.

16. Fukuzawa K, Hayashi T, Takahashi J, et al. Evaluation of six-point modi-


Conclusion fied dixon and magnetic resonance spectroscopy for fat quantification:
A fat–water–iron phantom study. Radiol Phys Technol 2017;10(3):
MGRE MRI is a noninvasive technique to evaluate FF and 349-358.
R2* value in BM of AL patients. It could be a powerful tool 17. Garbowski MW, Carpenter JP, Smith G, et al. Biopsy-based calibration
to evaluate BM abnormalities in children with AL. of T2* magnetic resonance for estimation of liver iron concentration

1362 Volume 58, No. 5


15222586, 2023, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/jmri.28755 by Manager Information Resources Bond University Library, Wiley Online Library on [14/12/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Wang et al.: Preliminary Study of Confounder-Corrected Fat Fraction

and comparison with R2Ferriscan. J Cardiovasc Magn Reson 2014; validation against MR spectroscopy. AJR Am J Roentgenol 2015;
16(1):40. 204(2):297-306.
18. Vogler JB 3rd, Murphy WA. Bone marrow imaging. Radiology 1988; 29. Lu WY, Zhao MF, Rajbhandary S, et al. Free iron catalyzes oxidative
168(3):679-693. damage to hematopoietic cells/mesenchymal stem cells in vitro and
suppresses hematopoiesis in iron overload patients. Eur J Haematol
19. Moafi A, Ziaie M, Abedi M, et al. The relationship between iron bone
2013;91(3):249-261.
marrow stores and response to treatment in pediatric acute lympho-
blastic leukemia. Medicine (Baltimore) 2017;96(44):e8511. 30. Chai X, Li DG, Cao XL, et al. ROS-mediated iron overload injures the
20. Arber DA, Orazi A, Robert Hasserjian R, et al. The 2016 revision to the hematopoiesis of bone marrow by damaging hematopoietic stem/ pro-
World Health Organization classification of myeloid neoplasms and genitor cells in mice. Sci Rep 2015;13(5):10181.
acute leukemia. Blood 2016;127(20):2391-2405. 31. Brissot E, Bernard DG, Loréal O, Brissot P, Troadec MB. Too much iron:
21. Lugauer F, Nickel D, Wetzl J, Kannengiesser SAR, Maier A, A masked foe for leukemias. Blood Rev 2020;39:100617.
Hornegger J. Robust spectral denoising for water-fat separation in 32. Lee JH, Park S. Differentiation of Schmorl nodes from bone metastases
magnetic resonance imaging. In: Navab N, Hornegger J, Wells W, of the spine: Use of apparent diffusion coefficient derived from DWI
Frangi A, editors. Medical Image Computing and Computer-Assisted and fat fraction derived from a Dixon sequence. AJR Am J Roentgenol
Intervention—MICCAI 2015, Vol 9350. Cham: Springer Lecture Notes 2019;213(5):W228-W235.
in Computer Science; 2015.
33. Yoo HJ, Hong SH, Kim DH, et al. Measurement of fat content in verte-
22. Zhong X, Nickel MD, Kannengiesser SAR, Dale BM, Kiefer B,
bral marrow using a modified dixon sequence to differentiate benign
Bashir MR. Liver fat quantification using a multi-step adaptive fitting
from malignant processes. J Magn Reson Imaging 2017;45(5):1534-
approach with multi-echo GRE imaging. Magn Reson Med 2014;72(5):
1544.
1353-1365.
34. Schmeel FC, Luetkens JA, Wagenhäuser PJ, et al. Proton density fat
23. Winfeld M, Ahlawat S, Safdar N, et al. Utilization of chemical shift MRI
fraction (PDFF) MRI for differentiation of benign and malignant verte-
in the diagnosis of disorders affecting pediatric bone marrow. Skeletal
bral lesions. Eur Radiol 2018;28(6):2397-2405.
Radiol 2016;45(9):1205-1212.

24. Chan BY, Gill KG, Rebsamen SL, et al. MR imaging of pediatric bone 35. Disler DG, McCauley TR, Ratner LM, et al. In-phase and out-of-phase
marrow. Radio Graphics 2016;36(6):1911-1930. MR imaging of bone marrow: Prediction of neoplasia based on the
detection of coexistent fat and water. AJR Am J Roentgenol 1997;
25. Vande Berg BC, Malghem J, Lecouvet FE, Maldague B. Classification 169(5):1439-1447.
and detection of bone marrow lesions with magnetic resonance imag-
ing. Skeletal Radiol 1998;27(10):529-545. 36. Konen E, Ghoti H, Goitein O, et al. No evidence for myocardial iron
overload in multitransfused patients with myelodysplastic syndrome
26. Mirowitz SA, Apicella P, Reinus WR, Hammerman AM. MR imaging of using cardiac magnetic resonance T2 technique. Am J Hematol 2007;
bone marrow lesions: Relativeconspicuousness on T1-weighted, fat- 82(11):1013-1016.
suppressed T2-weighted, and STIR images. AJR Am J Roentgenol
1994;162(1):215-221. 37. Kushner JP, Porter JP, Olivieri NF. Secondary iron overload. Hematol-
ogy Am Soc Hematol Educ Program 2001;2001:47-61.
27. Nguyen T-V, Melville A, Nath S, et al. Bone marrow recovery by mor-
phometry during induction chemotherapy for acute lymphoblastic leu- 38. Gandon Y, Olivié D, Guyader D, et al. Non-invasive assessment of
kemia in children. PLoS One 2015;10(5):e0126233. hepatic iron stores by MRI. Lancet 2004;363(9406):357-362.

28. Bashir MR, Zhong XD, Nickel MD, et al. Quantification of hepatic 39. Ramanathan G, Olynyk JK, Ferrari P. Diagnosing and preventing iron
steatosis with a multistep adaptive fitting MRI approach: Prospective overload. Hemodial Int 2017;21(Suppl 1):S58-S67.

November 2023 1363

You might also like